<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="hu-etal-2026-ci">
    <titleInfo>
        <title>CI-RELAY: Contextual Integrity Mediation for Privacy in LLM-based Multi-Agent Systems</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Runshan</namePart>
        <namePart type="family">Hu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yukun</namePart>
        <namePart type="family">Dong</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Jia</namePart>
        <namePart type="family">Bi</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Chenyang</namePart>
        <namePart type="family">Zhu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2026-10</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Hung-Yi</namePart>
            <namePart type="family">Lee</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Sunipa</namePart>
            <namePart type="family">Dev</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jessy</namePart>
            <namePart type="family">Li</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Desmond</namePart>
            <namePart type="family">Elliott</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Budapest, Hungary</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-462-0</identifier>
    </relatedItem>
    <abstract>Multi-agent systems powered by large language models (LLMs) increasingly decompose tasks by passing natural-language relays among specialized agents. This creates a boundary-level privacy failure: a non-entitled internal agent can learn private context before any final-output guard or tool-action checker observes a violation. We introduce contextual-integrity mediation for inter-agent communication, a formulation in which each candidate relay is treated as a commit that must satisfy recipient entitlement, attribute minimization, and transmission conditions before delivery. We present CI-Relay, a runtime mediation layer that induces CI scopes from runtime-visible norms and relay metadata, packages candidate messages as CI-labeled envelopes, projects recipient-scoped views, propagates transmission constraints, and records auditable mediation traces. On relay traces converted from three CI benchmarks, unmediated relays leak internally in 52.2% of cases. Relative to a metadata-matched LLM Guard Relay that receives the same induced scope, CI-Relay reduces internal leakage from 12.4% to 1.8% and downstream leakage from 10.9% to 5.6%, while improving utility and reducing over-redaction. Gains hold across three live LLM backends; multi-hop, prompt-robustness, and ablation studies show that the gains arise from structured CI mediation rather than generic privacy prompting, guard rewriting, or final-output checking.</abstract>
    <identifier type="citekey">hu-etal-2026-ci</identifier>
    <identifier type="doi">10.18653/2026.findings-emnlp.2421</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-emnlp.2421/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>51981</start>
            <end>52001</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T CI-RELAY: Contextual Integrity Mediation for Privacy in LLM-based Multi-Agent Systems
%A Hu, Runshan
%A Dong, Yukun
%A Bi, Jia
%A Zhu, Chenyang
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F hu-etal-2026-ci
%X Multi-agent systems powered by large language models (LLMs) increasingly decompose tasks by passing natural-language relays among specialized agents. This creates a boundary-level privacy failure: a non-entitled internal agent can learn private context before any final-output guard or tool-action checker observes a violation. We introduce contextual-integrity mediation for inter-agent communication, a formulation in which each candidate relay is treated as a commit that must satisfy recipient entitlement, attribute minimization, and transmission conditions before delivery. We present CI-Relay, a runtime mediation layer that induces CI scopes from runtime-visible norms and relay metadata, packages candidate messages as CI-labeled envelopes, projects recipient-scoped views, propagates transmission constraints, and records auditable mediation traces. On relay traces converted from three CI benchmarks, unmediated relays leak internally in 52.2% of cases. Relative to a metadata-matched LLM Guard Relay that receives the same induced scope, CI-Relay reduces internal leakage from 12.4% to 1.8% and downstream leakage from 10.9% to 5.6%, while improving utility and reducing over-redaction. Gains hold across three live LLM backends; multi-hop, prompt-robustness, and ablation studies show that the gains arise from structured CI mediation rather than generic privacy prompting, guard rewriting, or final-output checking.
%R 10.18653/2026.findings-emnlp.2421
%U https://aclanthology.org/2026.findings-emnlp.2421/
%U https://doi.org/10.18653/2026.findings-emnlp.2421
%P 51981-52001

Markdown (Informal)

[CI-RELAY: Contextual Integrity Mediation for Privacy in LLM-based Multi-Agent Systems](https://aclanthology.org/2026.findings-emnlp.2421/) (Hu et al., Findings 2026)

• CI-RELAY: Contextual Integrity Mediation for Privacy in LLM-based Multi-Agent Systems (Hu et al., Findings 2026)

ACL

• Runshan Hu, Yukun Dong, Jia Bi, and Chenyang Zhu. 2026. CI-RELAY: Contextual Integrity Mediation for Privacy in LLM-based Multi-Agent Systems. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 51981–52001, Budapest, Hungary. Association for Computational Linguistics.